Efficient Uncertainty Propagation in Gaussian Process-Based Extended Object Tracking
Eugen Ernst, Florian Pfaff · 2025
Extended Object Tracking (EOT) plays a crucial role in various applications within robotics, signal processing, and control, especially with advancements in sensing technology enabling multiple measurements per target. This paper proposes a novel EOT method based on Gaussian Processes (GPs), incorporating an extension for range measurements and explicitly accounting for uncertainties in Cartesian coordinates. We introduce an efficient method for the integration of uncertainty propagation. Empirical evaluations using Root Mean Square Error (RMSE), Chamfer Distance (CD), and mean Intersection-over-Union (mIoU) demonstrate the superior performance of our method compared to existing GP-based techniques. Additionally, we assess model credibility by comparing the Average Normalized Estimation Error Squared (ANEES) of the existing approach with our proposed method.